In 2015, Hans Strasburger and Bruno Waldvogel reported an unusual patient with dissociative identity disorder. After an accident she had been diagnosed with cortical blindness, and over years of therapy she began to recover sight in some identity states but not others. Visual evoked potentials, the cortex's electrical response to patterned stimuli, were absent in her blind states and present in her sighted ones. Switches between them sometimes took only minutes. A single visual system, it seemed, could be gated on or off depending on which self was presiding.
Cases like this unsettle a premise that has quietly structured consciousness science. The premise is that at any moment a brain hosts exactly one subject, whose experiences are bound into a single phenomenal field. The split-brain literature has long been the standard test of that premise, and its verdict remains contested. Pinto and colleagues' 2017 findings of unified responding in callosotomy patients reopened a question that Sperry and Gazzaniga seemed to have closed.
Callosotomy, though, is a rare surgical intervention. The more revealing question is whether unity fractures under conditions that require no scalpel at all. Dissociative identity, binocular rivalry, and hypnotic dissociation each probe this from a different angle. Read together, they suggest that phenomenal unity is neither a metaphysical given nor an illusion. It is an achievement: something the brain produces, maintains, and can, under specific pressures, partially withhold.
Dissociative Identity: Multiple Streams or Multiple Self-Models?
Dissociative identity disorder presents the most dramatic apparent fragmentation in clinical psychology. Patients show distinct identity states, each with its own voice, posture, autobiographical access, and reported amnesia for the others. Some patients also report co-consciousness, with one identity observing while another controls behaviour. For unity theorists, the question is whether these reports describe genuinely separate phenomenal streams or a single stream organised around shifting self-models.
Neuroimaging offers partial traction. Simone Reinders and colleagues used PET to compare neutral and trauma-related identity states in the same patients while they listened to autobiographical trauma scripts. They found state-dependent differences in regional cerebral blood flow and autonomic response. A 2012 follow-up showed that actors instructed to simulate identity states did not reproduce the clinical pattern. That result weakens the claim that DID is mere role enactment.
Behavioural memory studies complicate the picture. Rafaële Huntjens and colleagues tested interidentity amnesia directly by teaching material to one identity and probing another. The patients sincerely reported having no access to the material, yet their performance revealed transfer on implicit measures and, strikingly, on some explicit recognition tasks. Information was crossing the supposed boundary. What failed was the identity's ownership of that information, the felt sense that a memory belonged to me.
That dissociation between access and attribution is what a Dennettian reading predicts. In 1989, Dennett and Humphrey described the self as a centre of narrative gravity. On their account, DID may involve multiple narrative centres competing for a single body's report channels, rather than multiple independent streams. The brain stays largely integrated at the level of information processing. What has fractured is the self-model, the structure that tags contents as one's own.
The Strasburger case resists easy deflation, however. Absent evoked potentials point to gating at a relatively early cortical stage, not merely in after-the-fact narrative. The most defensible conclusion is that DID reveals disunity of the self-model, along with partial, state-dependent gating of access. Whether two phenomenal fields ever coexist remains open. Claims of co-consciousness have scarcely been tested empirically, and that gap is itself striking.
TakeawayDID suggests the self can fragment while the underlying information stays shared. If so, 'one subject' and 'one stream' are separate claims that we have long treated as one.
Binocular Rivalry: Unity as Active Enforcement
Binocular rivalry offers the opposite kind of evidence: not fragmentation under pathology, but unity actively enforced in healthy vision. Show a face to one eye and a house to the other, and observers do not see a fused hybrid. Perception alternates every few seconds, with one image dominating while the other is suppressed. Both retinal signals remain continuously available to the visual system throughout.
Nikos Logothetis and colleagues traced where this competition is resolved by recording from macaques trained to report their percept. In V1 and V2, only a minority of neurons changed their activity with the perceptual alternation, and the proportion rose in V4 and MT. In inferotemporal cortex, Sheinberg and Logothetis found that roughly 90 percent of responsive cells tracked what the animal reported seeing. Frank Tong and colleagues found the human analogue with fMRI. Activity in the fusiform face area and parahippocampal place area followed the percept, not the stimulus.
Rivalry therefore seems to reveal a strong unity constraint: two incompatible interpretations of the same location cannot coexist in experience. The constraint is less absolute than this suggests, though. Piecemeal rivalry produces patchwork percepts, with fragments of each image combining across the visual field. Kovács and colleagues split complementary pieces of two images between the eyes, and observers saw coherent wholes assembled across them. The brain rivals over interpretations, not over eyes.
Suppressed stimuli also remain causally active. Tsuchiya and Koch developed continuous flash suppression, which keeps an image out of awareness for seconds or longer. Using it, researchers have documented priming, attentional orienting, and adaptation effects from images subjects never consciously see. Content excluded from the unified field is not deleted. It continues to shape the processing beneath it. Researchers still argue over how large such unconscious effects are, but their existence is reasonably secure.
No-report paradigms add a methodological caution. Frässle and colleagues tracked rivalry through eye movements and pupil size rather than button presses. Much of the frontal activation previously attributed to conscious switching diminished. Rivalry, then, reveals a unity enforced at the level of a coherent world-model, operating over massively parallel processing. Unity looks less like a container and more like a selection policy, often resolved below the level of frontal report systems.
TakeawayRivalry suggests unity constrains interpretation rather than information. The brain can carry two incompatible signals at once, but it lets only one of them become the world.
Hypnotic Dissociation: Division in Self-Knowledge
Hypnosis offers a third angle: divisions that can be induced experimentally in otherwise healthy, highly suggestible people. Ernest Hilgard's hidden observer studies of the 1970s became the canonical demonstration. Subjects given hypnotic analgesia reported minimal pain while their hands were immersed in ice water. When cued to let a hidden part speak, they reported pain intensities closer to waking levels, through automatic writing or key presses.
Hilgard's neodissociation theory read this as evidence of parallel cognitive systems under a weakened executive. On that view, a concealed stream registered what the primary stream denied. Nicholas Spanos offered a pointed sociocognitive critique. By varying instructions, his group showed that hidden observers reported more or less pain depending on what subjects were told the hidden part would experience. A genuinely independent stream should not bend to expectation this way.
Contemporary theory has largely relocated the division. Dienes and Perner's cold control theory proposes that hypnotic responses are intentional actions carried out without the higher-order thought that one is intending them. The action is real and goal-directed. What is missing is an accurate metacognitive representation of where it comes from. This fits subjects' reports that a suggested arm levitation simply happens. Consistent with some data, the theory also predicts weaker metacognitive monitoring in highly hypnotisable people in certain domains.
Neural evidence shows that suggestion has real effects on processing. Amir Raz and colleagues gave highly suggestible subjects a posthypnotic suggestion to see words as meaningless symbols. Stroop interference was substantially reduced or eliminated, and anterior cingulate activity changed alongside it. Stephen Kosslyn and colleagues found that suggestions to see colour in grayscale images shifted fusiform activity in the suggested direction. Suggestion can reach surprisingly deep into processing usually considered automatic.
For unity theories, hypnosis is instructive precisely because it does not deliver two phenomenal streams. It delivers a dissociation between first-order processing and the higher-order representation of that processing. The division lies within the architecture of access and ownership. That makes hypnosis a natural testbed for higher-order and global workspace accounts. It also poses a quiet problem for any theory that equates unity with the bare co-presence of contents.
TakeawayMuch of what feels like a division in consciousness may be a division in self-knowledge: the mind doing something without knowing that it is the one doing it.
Across these three domains, a pattern emerges. DID fractures the self-model while information leaks across identities. Rivalry enforces a single coherent interpretation over parallel processing that continues regardless. Hypnosis separates action from metacognitive ownership. None cleanly demonstrates two simultaneous phenomenal fields. All three show that unity is produced by identifiable mechanisms that can be modulated.
For theory, this favours frameworks that distinguish phenomenal unity, access unity, and subject unity rather than treating them as one property. Integrated information theory, global workspace theory, and higher-order accounts make different predictions about which of these should break first. These cases supply the data needed to test them.
For research, two priorities stand out: applying no-report methods to dissociative states, and rigorously testing claims of co-consciousness. Whether a brain can host two subjects at once remains an open question. What is giving way is the assumption that it obviously cannot.